Bring things spring from small seeds. This is the thought that keeps running through my mind when I think of Artificial Consciousness in computers. Ever since I saw the Imitation Game and the story of Alan Turing, I wondered how such an intelligent man could think that computers could think. Of course he stipulated that the thought process was different, than in humans.
Then along came Dr. Stephen Thaler who artificially introduced the idea of perturbations in computer "thinking" and got a patent for it. A perturbation is essential to artificial consciousness. Essentially, a computer is programmed to linearly follow an execution path of its program. Even in artificial neural networks, the output of one perpceptron is fed into another layer. It is a linear defined path. Thaler introduced perturbations by selectively killing off perceptrons in a layer and in what he describes as artificial neuronal near-death, and the machine becomes a creative design machine. His landmark example was a neural network that identified coffee cups and when it was brain-damaged, it came up with creative coffee cup designs.
Perturbations can come from all things. They can come from random events. But in the state of consciousness of any sentient thing or being (notice I now have to add things, because computers have the ability to become sentient), perturbations can come from the state of consciousness itself. A prime example is worry. We observe something within our conscious sphere, and we think about it, make a judgement about it, add the judgement to the thought process, and keep recycling the thought in an obsessive compulsive state, and you have what is known as worry.
Going back to the opening statements of big things spring from small seeds, the thought struck me that I could make my computer worry. It would be a small worry program to start with, but then I could hop it up to another layer of abstraction and make a universal computer worry module that could become part of the Operating System. It would be the Worry Service.
Here's how a simple version would work. The Worry Service runs a task manager, a memory monitor or a CPU usage monitor in the background. The minute that it detects that memory or CPU is approaching 100% or saturation, it kicks the worry.exe module. The worry module essentially assumes the highest thread priority, prints out on any display saying "I am incredibly busy" and deallocates processing priority to the heavy task, slowing it down. It then detects that the task is slowed down, and kicks another worry module about its lack of performance. The worry modules are able to be queried, and their response is always "I am incredibly busy and it is affecting my performance". The worry module also writes to every log that it can, and using a machine learning neural network, reinforces the worry parameters so they automatically fire at lower thresholds. Once the busy task is completed, the worry abates and slows down, and the computer becomes efficient again. Of course, if the machine-learning neural network is too effective and eager in kicking in the worry, it becomes a compulsive worry, and needs to see a programming computer psychiatrist to up the thresholds of its worry mechanism by running a few positive reinforcement training epochs. All of this technology is available now.
But I can just see it. Some Goth programmer will chain all of the worry modules into the depression module, making the computer virtually worthless for sustained work.
The very first step to scary artificial intelligence, is making a computer with the ability to navel gaze. This is a start. I am convinced that human consciousness is merely an accident of an over-developed tropism, and the evolution of Artificial Consciousness can start with this simple step -- a computer worry wart. Windows machines will be the worst worry warts and the most depressive among the conscious computers.
We have a new thing to worry about in the skies. In the 1950's, it was airline crashes because metal fatigue and cabin pressurization was not that well understood. In the 1960's, we had the airline celebrity extirpation phase, taking out stars like Buddy Holly, Richie Valens, the Big Bopper, Patsy Cline, Jim Reeves, Otis Redding and boxer Rocky Marciano to boot. In the 1970's we started having hijackings to Cuba, and terrorists attacks that continued until the present day. We have had underwear bombers, crashing planes into skyscrapers, bombs R us, ground to air missiles and all sorts of imaginative ways to take airliners out of the sky. And now we have a new threat in the skies -- The Pilot Going Nuts Syndrome. It probably happened on Malaysian Airlines MH370 and we have the GermanWings pilot deliberately crashing into the Alps.
The Pilot-Going-Nuts Syndrome is totally stoppable with Machine Learning, Artificial Intelligence and a bit of remote control.
For example, we now have drones taking off from Alabama, flying to the Middle East, blasting a terrorist in his tent upwards to meet his Allah in pieces and collect his virgins in some godforsaken place and then the drone flies home while the operator is eating a pulled pork sandwich somewhere in a bunker near Huntsville (irony in the name as well). So the ability for remote control is well established.
Now let's take machine learning. After an airplane flies a route from Barcelona to Duesseldorf ten times, any machine-learning program knows the flight plan by rote. Even Microsoft's Azure platform in the throes of the Blue Screen of Death, is smart enough to learn that route. So if you embed a program like that into the avionics, and you input the flight plan in as well, any Pilot-Going-Nuts sufferer could be thwarted.
The minute that Co-Pilot Cuckoo For CocoaPuffs tries to take the plane off autopilot in direct deviation of the flight plan (especially at cruising altitude), the smart avionics notifies Air Traffic Control and asks for a OK semaphore. In the meantime, the computer says to Lieutenant CocoaPuffs "I'm sorry Dave, I can't let you do that". In case of real emergency, any major deviation to the flight plan could be done using biometric authentication of both pilots like fingerprints or iris scan. Having both pilots go crazy is a huge longshot, unless their turbans are made from the ISIS flag. The double authentication is that if one pilot uses the loo, the plane and passengers are not sh*t-out-of-luck if the other one takes leave of his senses.
This is so easy to fix, I don't know why its not a slam dunk. The only reason that I can think of, is that taking a chance with batsh*t crazy pilots is a lot cheaper than refitting aircraft with anti-crazy avionics. However the system will pay for itself by preventing just two airplane annihilations by apesh*t crazy, maniacal sky jockeys.
Self-driving cars are first, and self-flying planes will be next. And twenty years later, we will even have self-cleaning toilets. It will follow the same trajectory as we put a man on the moon in 1969, and luggage never got wheels until the late 1980's. In the meantime, tonight I am painting huge Rorschach ink blots on my carry-on luggage, and if any of the flight crew gets googley-eyed or starts to salivate when seeing my luggage, I am getting off. The momentary upgrade to first class when the plane hits the mountain is not worth it.
Those b*st*rds are going too far. Even though I am a data miner, I have a great concern as a data privacy advocate. Essentially Twitter & IBM are teaming up to mine your Twitter Stream to monetize your posts. They will take your tweets and try to sell crap to you, or worse, sell your data to other companies.
Here's how it will work. If you post that your mother died, you will see a crematorium or undertaking ads. Tweet about spending some time in the hospital, and you might pay a higher health insurance premium because they will sell that info to insurance companies. The same about driving fast. Tweet about your kid going to college, and you will get a full court press on everything from college choices to clothes for university life.
It sucks. It just isn't right. You have three choices. You can vote with your feet and leave Twitter. I have already left Facebook and LinkedIn. Twitter is my last stand.
You can carry on, but in a previous blog post, I mentioned that the most dangerous thing about Big Data Mining, is that data mining can make assumptions about you that simply aren't true, and you may be categorized into a list that you don't want to be on. It could affect your job, your security clearance, your credit score or who knows what.
You could self-censor, but censorship is wrong, even self-censoring.
I like the last option - f*ck with the machine learning, and deep learning and data-mining. How? Obfuscate. Here are a few things that I will do.
1) Disable all location services for tweets.
2) Disable all location services that your smart phone takes. It writes the location into the EXIF data. It also writes date and time and camera type, etc.
3) Google for a free EXIF editor, and remove all EXIF data from your pics.
4) Do not put your actual location in your bio. For example, I follow a dude, who's location is : Where I Have To Be
5) Put in a fake town where you live. If you have a dog named Rover, put down that you live in Roverville. You can still keep your same state.
6) Never use your middle name or initial. It's just one more authentication factor.
7) When social media streams are mined using NLP or Natural Language Processing, an important part of that is finding "possessive determiners". Don't use them. Possessive Determiners are words like my, your, her, etc. If you tweet "Its my birthday", even the dumbest NLP data mining machine can pick it up. However if you say "Welcome to Birthdayville, Population Me", not even the smartest NLP machine can pick that up. Get rid of possessive determiners in your Tweets.
8) Practice Typoglycemia. http://en.wikipedia.org/wiki/Typoglycemia Here is an example that would totally screw up a deep learning machine:
"I cdn'uolt blveiee taht I cluod aulaclty uesdnatnrd waht I was rdanieg: the phaonmneel pweor of the hmuan mnid. Aoccdrnig to a rseearch taem at Cmabrigde Uinervtisy, it deosn't mttaer in waht oredr the ltteers in a wrod are, the olny iprmoatnt tihng is taht the frist and lsat ltteer be in the rghit pclae. The rset can be a taotl mses and you can sitll raed it wouthit a porbelm. Tihs is bcuseae the huamn mnid deos not raed ervey lteter by istlef, but the wrod as a wlohe. Scuh a cdonition is arppoiatrely cllaed Typoglycemia .
"Amzanig huh? Yaeh and you awlyas thguoht slpeling was ipmorantt."
9) User slang. If your gas pedal foot itches to drive a BMW, call it a beamer or a beemer and don't capitalize the word.
10) Use alternate spelling. Ime a bygg phan of Neel Yoongs mewsic.
11) Throw in rand o m s pac es in yo ur sente nce. Or e*ven the od*d star will do.
12) Never tweet your age, your spouse or partner (I see married to @sweetiePie all the time) or any other information. It is okay to list your employment of academic institution and that leave a lot of room to fool the NLP machines if you work at the Big Blue, or teach @ the Yard (thanks to the Harvard profs that follow me -- appreciate it).
Using these simple tips will cause the data mining and perceptrons scanning your feed to take a pass on what you type. Now is the time to bowdlerize or obfuscate your account.
I think that the bigger answer, is to startup a new hybrid of Twitter and Facebook that guarantees information privacy. But in the meantime, let's be careful out there as to what we post. And remember, its not that difficult to deke out smart machines.
(click on pic for larger image)
Analytics and Big Data are not always the panacea to solve everyday problems in life. This is especially true in our business. We are remarketers of automobiles. We started from a large bricks & mortar auto auction on the East Coast, and we wanted to transfer our business to the canvas of the web, and bring incredible added value to a necessary, but un-glamorous industry with a technology platform that is 21rst century instead of the 18th century auction. And the foundation patents that we have pending in the private buyers network prove that we have done our homework and done technological wonders by introducing private buyers networks; auto-escalation to buying groups; the timed auction where the new car dealership that is dealing the trade is now the auctioneer; as well as robot bidders; and a whole raft of features including an onboarding mobile app that makes loading a car onto our platform a breeze.
But there was one more nut to crack, and that was valuations. You see, we are information brokers at the heart of it. We level the playing field between buyer and seller so that the seller gets a fair price and the buyer (who is usually a second-hand car dealer) gets a piece of decent inventory to make a healthy profit. Ergo we are the one that should provide a valuation to grease the wheels. Making money in remarketing automobiles should not be a zero sum game.
Black Book, Kelly, Blue Book, VAuto and all of the other valuators in the marketspace use analytics to determine guidelines for fair valuations. However, standard statistical practices aren't good enough for several reasons. For example, Black Book collects reams and reams of data from car auctions around the continent, and uses their statistical tools to come up with a published valuation that is sold to dealers everywhere. While one may be confident that it is indeed a standardized valuation (approximately) should that car go to auction, but there is a little secret in the auto business. Up two-thirds of the trade-in vehicles never hit the auction. Why? Because of relationship selling! People buy and sell from people with whom they have done business before and trust. The chain goes like this. When a trade-in comes in that is not suitable for the lot, the used car manager gets on the phone to his go-to guys. If that doesn't work, he taps a few wholesalers that he knows. Failing that, a car goes to auction. That is why auction prices are skewed. They do not truly reflect the value of the car. Either they are under-valued due to poor bidding at a particular auction or they are over-priced in a spate of auction fever. They are not the same price as one would get from relationship-base wholesaling. Auto auctions are the third step in the remarketing of a trade-in. Everyone else thinks that it is the first step in the process. We know better.
The sad fact in the business, is that if an auto sits for any length of time on a dealer's lot, he is losing money on it. Most used cars do not end up on the used lot at the dealership where they were traded.
When you walk into a new car dealer with a trade-in, you are giving him a problem right away. He doesn't really know what your car is worth. He may have a good feeling. He will make you an offer, but ultimately whether the trade-in comes close to his valuation or not, and whether he makes money on the trade, is a crapshoot.
Analytics in automobile valuations fail for several reasons. The chief one is that analytics relies on one right answer when presented with a pile of variables. You can analyze all of the big data from every single auction, and you would be hard-pressed to come within a nominal number that is within, say 10% of what a car will sell for. The same car with the same mileage will have a different valuation and sale price for every auction that it goes to. That is the key. What a car sells for, is what it someone is willing to pay for it at that given time! And you can take into accounts brands, defect history, buying patterns, locale, color, mileage etc etc whatever, the two bottom-line parameters in any used car sales and valuations are unknown to the buyer. They are (1) how the car was driven over its lifetime and (2) how it was maintained. These factors transcend all brand, mileage and other data points. And the buyer has to divine the answers by looking at the car or consulting an Ouija board.
Click on info-graph above. It is a view of the "Loudspeaker graph of valuations". Analytics and big data will give us an average value for make, year, model, options, mileage and condition. That is the oblong rectangle in the middle. If a trade-in is less than 4 years old and has low mileage, then valuating the car is a slam dunk. The junior guy on the lot who still doesn't have to shave the fuzz off his face every day can do it. But as a car increases in age, the factors that go into valuations start to multiply. Was it driven by a little old lady to bank every Friday and idled while her gang robbed it, and parked for the rest of the week? Was it driven by a soccer mom, who had practice in the next town sixty miles away every day? The green X can demonstrate a high range of the valuation and the pink X shows the low range of valuations. Put simply, a valuation is a probability of what someone will pay for the vehicle on a wholesale level.
So how did we crack the valuations conundrum? Well we do have a machine-learning, artificial neural network valuator, but we haven't put it into production yet. It simply is not ready at this point. But we have put into production another source that is more accurate than Black Book, more accurate than any auction data, and actually intimates what a particular person in a particular locale will pay. How did we do this? We went to crowdsourcing. The crowd is always right!
Local secondhand car dealers have intimate knowledge of their terroir. The secondhand car dealer who buys the trade-in, has intense knowledge of the very local market. He knows what will sell and what won't, and for how much. He knows what kind of car they need as a lure on their lot to generate foot traffic. He know what sells quickly and what doesn't. He knows who buys what in his neighborhood. He knows that when the youngish, single female in her very early twenties is looking for a car, that red Chevy Cavalier in the back is just the ticket. Their livelihood depends on it. So we tap a bunch of them with our technology.
We have developed a crowd-source, social network, exclusive zone tool to give local, accurate appraisals using the latest in our platform technology. It's a win-win situation. The new car dealership uses a mobile phone to scan and explode the VIN number so there is no typing of the VIN, and instantly all of the data about that specific car appears, delivered from our platform. Then a quick, visually-based condition report is ticked in, and sent to the dealer's professional social network or his exclusive zone, based on relationship wholesaling. The receiver, the guy who gets the opportunity to valuate the vehicle, is in a prestigious position. He sees upcoming inventory before it goes to auction, and he gets first dibs over the rabble on the good stuff coming up. With his valuation, he can put in an offer. But we even kicked that up a notch. We have created the day trader of automobiles. One of members of the exclusive zone, can flip it to his group either to find out what it is worth, or to gauge interest in someone buying it.
If the car is somewhere in the bottom cone of the loudspeaker in the above diagram, and everyone takes a pass, then our analytics kick in. We don't try to valuate it there, but we find buyers who have bought this kind of car before. After a few hours, the system sends the car to these guys for offers. If that doesn't work? Then the car goes to auction on our platform. And if that doesn't work, it goes to a classified type of wholesale listing. All of this happens without a human being present. The platform does it all. We will sell the car no matter what, and we will sell it for what its really worth!
So, analytics and Big Data may fail in the valuation, but it doesn't fail us in find a buyer. We sell used cars, but with the tech-infused platform that we have, with very unique Intellectual Property, I really don't mind being called a used car salesman.
The one thing that I have learned from designing a remarketing platform, is that there is a market for absolutely everything. This was exemplified by my trip to Siberia. Near Finland's border I discovered a Russian millionaire who made his money from crap. Literally. He was a chicken farmer. Up on the north peninsula to Finland, the ground is rocky and devoid of nutrients. Chicken crap is an amazing fertilizer full of nitrogen. He would trade piles of chicken crap for used Swedish cars - mostly beaters that the Finnish farmers bought cheaply. Getting a car in Russia is tough if you are not in a major center, and he marked up his cars considerable and within a few years, became a millionaire. Like I say, there is a market for everything.
I was watching a video on the modern fur trade, from the trapping right down to the auction. It was fascinating because when we think of fur trade, we think of glamorous women in fur coats on the runway. In the case of furs, there is an industrial market for crap furs. The Chinese are the biggest buyers. They will buy rabbit fur, and furs that are not good enough for clothing, and use them for toys, novelties, lining on the inside of boots -- wherever.
So there is a job description in the fur purveying process called a fur grader. Before the auction, he and his staff go through the bales and bundles of furs, and groups them according to quality and type. Bales for sale consist of similar types and grades so that the quality of the bale is consistent.
As I was watching this, it struck me that the same thing will happen to data. There will be a data grader who will assemble datasets, grade them, evaluate their marketability, cleanse the data, and put it on a data exchange. In the case of valuable datasets, there will be a data auction. And if you are thinking of starting a data exchange, have I got the platform for you !!!
You will also get the arbitrageur and day trader of data. If I see I dataset that is going cheap, I just may pick it up. You see, I will have machine learning on my side, so I can process it. I will do goal-oriented mining, and create multiple datasets of differing things with differing values. Much like capital partners buy and ailing company, and break up and sell its parts for more than the overall thing was worth, the same will be done for data.
It's a brave new world out there. Folks in the industry used to think that the never-ending, endlessly multiplying streams of data were a scourge. They are actually a raw material and an asset. After all, there is a market for everything. Sign up for my emails on the right for more.
# DO-IT #tag for NLP machine learning for this article.
There is a shocking liberty and data privacy incident going on in Canada. The Canada Border Services or Custom Guards stopped a traveler and asked him for his smart phone password. He refused. The traveler was charged with obstructing a customs officer. In Canada, a smart phone can be considered like any other of your belongings and liable to be searched. The constitutionality is being tested in May, but the border authority still asserts the right to check your smart phone, tablet or computer. Here is the link:
So, if you want to be immune from these sorts of fishing exercises what can you do? Here are some tips:
1) Offload documents to a book type external disk. A terabyte drive is less than $100 now. Documents that you will need can be stored in the cloud. There are several cloud providers for file services.
2) Carry documents on a USB stick on a key chain. There are several USB key chain novel items that do not even look like USB keys. Put it on a keychain in plain site. Or here is a pair of USB keys that are earrings:
3) Offload your photos to other storage. They may want to clone your photos to see if you are lying about where you traveled to.
4) When traveling, never use your mail program like OutLook that resides on your computer. Just by firing it up, one can see all of your contacts.
5) If you do have an email account that doesn't have a web interface or browser interface, create a gmail account, that is accessible by browser, and for the duration of travel, forward your mail to the gmail account.
6) Do not download the mobile app for email, either Yahoo, or Gmail or whatever. Always use the browser.
7) Before crossing international borders, always erase your browsing history and delete all of your cookies. That way, it will not even be apparent that you have a web email account.
8) It goes without saying, do not have questionable documents or pictures on your devices. You know what they are.
9) In many countries, your hard disk is surreptitiously cloned (notably China and Israel). So even if you delete documents, all that is deleted is the memory reference to them. They can be forensically reconstructed. The solution is that sensitive documents are never written to disk. They are copied to a USB stick, and edited on the stick. That way, temp edit files that are created when you open the document, are not written to the disk, but in the same directory on the USB stick. If the system doesn't clean them out (and they do stick around), they will not be on your cloned disk.
10) Your smart phone is your life. It is the repository of who you are. Giving up the password is opening the book on your life, your finances, your business, everything. If you are really concerned about this, the solution is to buy a cheap flip phone while travelling. Remove the SIM card from your smart phone, and put it in the cheap flip phone. You still have conventional SMS texts, phone and a browser for you email, but you don't carry your own personal data repository around with you.
11) Never use free airport WIFI. Always use your 3G or 4G data in the airport. All of the intelligence agencies in the world listen in, (and so do I when I am bored). I just fire up my network monitoring tools and watch the data go by.
12) Finally, if you are a bona fide company, or a High Net Worth Individual looking for an enterprise or robust solution to the empty laptop, send me an email DataPrivacy-at-mail.com (substitute "-at-" with "@") We have an enterprise, secure solution where the data is safely stored in a bunker in the Bahamas, and access is through a hardware key to your computer with intense SSH/SSL encryption and tunneling. Be advised though that we do due diligence and KYC (Know Your Customer) because we want purely legitimate business with privacy concerns. Our usual customers are financial institutions and multi-national or international corporations operating from a G20 country.
The age of information really erodes personal privacy, but there can be technology solutions as well.
I really don't understand why houses haven't evolved with the knowledge of science. Our habitat, our dwelling is where we spend the most time, and yet we live in a box-like structure that is nothing more than a synthetic cave.
Anyone who peruses YouTube can see folks building rocket mass heaters that heat houses with minimal amounts of fuel, and store the heat like a thermal capacitor in a big rock or concrete mass in the house that radiates back its heat all day.
Well I have a better idea base on Nature or bio mimicry. Everyone schooled in physics knows that water stores heat better than rock or concrete. Suppose you had a huge mass of water in your house, say a sealed plastic reservoir with an anti-freeze and an anti-fungal agent. This mass of liquid would be a circulatory system for the house or dwelling. If you needed heat, it would not only collect the heat from a heat source like a furnace or a stove, but the reservoir would be connected to solar collectors. The liquid circulatory system would spread the heat around the house.
The thermal mass capacitor would cool in the summer and heat in the winter using heat pump technology.
Each house would also have a ambient heat conduit that would force the warm air at the top of the room near the ceiling to the floor. These could be passive or solar operated.
This business about air exchange baffles me as well. I don't understand why houses do not have a ventilation system where the entire volume of the air in the house is exchanged on a daily basis. In the winter the heat from the air could be recovered with a heat exchanger, and the reverse with air conditioning would take place in the winter. The air would be HEPA-filtered to remove pollen, dust and allergens.
I don't understand why every surface pointing to the sun is not covered with solar energy converters. Every little bit counts.
I don't understand why high traffic areas do not generate electricity in a house with the piezo effect by people walking over the floor, or for that matter, driving in front of the house, or walking on the sidewalk in front of the house.
I don't see why windows can't be covered at night with automated shutters to preserve heat or air conditioning.
I don't see why the heat cannot be recovered from bath water going down the sewers. I don't see why houses don't have passive reverse osmosis systems to obviate the need for plastic bottled drinking water.
I don't understand why houses do not have built-in composters to handle wet garbage, and instead of garbage collection, have compost collection. Surely someone can advance the art of composting to reduce the carbon footprint of garbage pick-up.
There is lots of room for improvement, and the technology is cheap enough now to implement. We need a massive education for the architects, city planners, and home buyers to re-evaluate what they should be looking for in a house.
I am having a huge problem with Cordova/Phonegap and iOS. The app is supposed to take a pic and upload it to the server. It works perfectly on the iPad but fails miserably on the iPhone. It should work on both.
Here is the error that is thrown:
Plugin 'CDVCamera' has a pending operation, memory purge is delayed for didReceiveMemoryWarning.
So I googled the error. There are only about four entries and none of them helpful. One of them was Chinese. So I popped the answer into Google translate, and this is what I got:
Oh, It is what iOS7, or Will not tried in another device?
Do Will does not implement the functions except after camera function? , May now once please confirm there is a heavy processing such as image processing in other So, I'm Although it is the main issue is the iPad mini instead of iPad mini2?And I totally bring up the story of another question, but the contents of the current "north of the traveler," Mr. during the inquiry, when you change from β version to the official version, memory release of native component has not been correctly, memory may have adapted to generate it by tight is very high I want is you answer with. Both devices were asked earlier I will remember it was a memory-mounted 512mb. And diverted the answer in another thread, but has a large memory consumption of iOS7, you fell as soon as monaca is not the chitin and memory management. Here How do leading to talk of memory management shortage at the time of the previous official version migration I guess hands of one that you wait for the after performing the memory released monaca side?
I would definitely call this a fail. In the meantime, if anyone knows why the iOS Cordova framework is throwing this error, please let me know. Thanks.